Download Conditions on Clitic Doubling - University of the Basque Country

.Conditions on Clitic Doubling:
The Agreement Hypothesis
JONFRANCO
(University of Southern California)
o. Introduction
If we define agreement as a category X that shares <p features with a category Y
in a linking relation, it could be said that Spanish object clitics and their doubles
constitute an agreement system. In this paper, I claim that this agreement system
does not run independently in the grammar of Spanish as a .supplementary system,
on the contrary, it is part of a more general relation, that is, argument verb agreement. Thus, the label Agreement Hypothesis, stands for an analysis that equates the
traditionally called object clitics with subject agreement morphemes.
Any analysis that considers object clitics in Spanish as agreement morphemes,
such as Borer's (1984) and Sufier's (1988), must answer the following questions: (1)
Why is object agreement not always obligatory like subject agreement is? and (11)
why is direct object clitic doubling so much constrained? I propose that a morphosyntactic theory of Agreement built on the data from Spanish and also on a comparison of Spanish object agreement with agreement patterns found in other languages
provides a straightforward answer to those questions.
.
In Spanish, object personal pronouns of' first, second and third person must
undergo object-verb agreement obligatorily, as in (1). It "is therefore, the non-obligatoriness (or sometimes the impossibility) of object-verb agreement with third person
NPs that prevents Spanish from exhibitit;1g"a sheer ~bject agreement pattern like the
one existing in languages like Basque, as illustrated in (2) and (3):
(1)
Juan no *(loi)
escucha
Juan not him-CL
listen'
Juan never listens to him.
a eh nunca.
to him never.
* Early versions of this ,paper 'Yere presented at the Third Student Conference in Linguistics thac took place in
MIT on February Q-10~ 1991, the XxI Linguistic Symposium on Romance Languages held at DC Santa Barbara on
February 21-24, 1991 'and the Summer Workshop in Linguistics hosted by the Universidad de Deusto in July 1991.
Without doubt, this ne~ version has profited a lot from such great audiences. I am also grateful to Mario Saltarelli
for fruitful discussion, help and encouragement throughout the process of w~iting this pape~. "I am indebted as well
to )oseph Aoun, Alazne Landa, Carmen Silva-CorvaIan, Enric Vallduvl and Marfa Luisa Zubizarreta for their
valuable comments upon its conception.
[ASjU, XXVII-I, 1993.285-298]
286
JONFRANCO
nunca.
_never
(2)
Juan no (loi)
escucha
Juan not him-CL
listen
Juan never listens to Pedro.
(3)
Guki liburuaj
]onik
erosi dj- iOk- gUi.
We-Erg book-Abs - Jon-Dat buy Abs-3Dat-lErg
We have boughtJon the book.
a Pedroi
to Pedro
On tpe benefit of the Agreement Hypothesis, I first claim that Spanish DO-NPs,
by virtue of their being third persons, may take a zero object-verbal agreement
morpheme, patterning in this way with other third person verbal agreement morphemes in otherpro-drop languages, such as subject verbal agreement morphemes in
Basque and Irish, and DO verbal agreement morphemes in Jacaltec, Cuzco Quechua,
etc. For example, let us look at (4) and (5) as instances of the latter:
(4)
(5)
J acaltec (Craig 1977)
a. Hayin x- 0- w- ute hun-tie
I Asp-3Abs-1Erg-do one-this
I did this.
b. Ch -in haw -ila.
Asp-lAbs 2Erg-see
You saw me.
Cuzco Quechua (Ortiz de Urbina 1989)
a. Riku- 0- ni Juan.
b. Riku-wa- n.
see -3Acc I Juan
see -lAcc 3Nom
I see Juan.
He sees me.
In (4a), the third person Absolutive NP does not have any corresponding phonological agreement marker in the verbal inflexion. This -phenomenon, although typical
of Ergative languages is also manifested in Nominative/Accusative languages, as
illustrated in (5~) by Cuzco QQechua.
1. Clitics as verb-object agreement morphemes: basic considerations
Everett (1989) establishes four basic types of clitics, as shown in (6):
(6)
Everett (1989): Clitic Parametrization
A. [+Case, +Argument]: requires thematic role and does not allow
clitic doubling.
B. [ + Case, -Argument]: allows clitic doubling (Yagua, Rioplatense
Spanish)
C. [-Case, +Argument]: does not allow clitic doubling (eg. English
passive morpheme: -en)
D. [-Case, -Argument]: these clitics have agreement and can
identify empty categories (eg. Subject
agreement morpheme)
As for Spanish, the choice is between Band D, since all dialects of Spanish allow
clitic doubling to a certain extent. A and C classify clitics as arguments. However,
for objects clitics to obtain argumenthood they would have to occupy a canonical
CONDITIONS ON CLITIC DOUBLING: THE AGREEMENT HYPOTHESIS
287
object position at D-structure, which would imply that there has been subsequently
clitic movement in order to yield the surface representation in (1) and (2). Strozer
(1976) and Rivas (1977), within the transformational grammar framework, already
proved that a movement analysis turns out to be problematic for sentences like (2)
above. To put it simply, if Spanish clitics were originated in an argument position:as
heads and then cliticized to the verb, the doubled constituent would wrong~y
occupy the position of the trace left by the clitic. Thus, the structure-preserving
principles on movement automatically discard A or C as candidates to categorize
Spanish clitics. In the next section, I will attempt to show (a) the advantages of the
Agreement Hypothesis over other hypotheses that consider clitics as unstressed
pronouns marked with Case, and (b) the problems that an analysis that places
Spanish under B may run into.
1.1. Simplification of Case Assignment
To my knowledge, there is no test that discriminates clitics. either as B or D.
However, Casewise, the classification of Spanish clitics as object-verb agreement
morphemes, gives us a clear advantage in the grammatical description of Spanish,
since the clitic and its double do not have to compete to receive Case in structures
like (1) and (2). Consequently, we could get rid of the machinery of Case absorption
and Case transmission put forward in ]aeggli (1982, 1986) and the subsequent
problems that this theory entails, as argued in Sufler (1988).
In addition, since structural Case is assigned by lexical heads from left to right,
Spanish being a head initial language, if we categorized object clitics as B [+Case,
-Argument], the directionality of this type of Case assignment under strict government would have to be reversed for these items, unless one assumes that clitics· are
generated post-verbally, an analysis that, again, we want to avoid for the reasons
stated in section 1.
1.2. Historical change in Spanish clitics
Although in Old Spanish some clitics could form chains with prepositional
phrases, hi 'in that' and en 'of that', in today's Spanish clitics only form nominal
chains. This could be an indication of the fact that the clitic system has developed
into a paradigm of object agreement, as pointed out by Silva-Corvahln (198l).
Probably, the most outstanding difference existing between clitics in Old Spanish and clitics in today's Spanish is their position in the sentence which points
directly to a possible change in their syntactic status. Whereas in present-day
Spanish, clitics are restricted to preverbal or postverbal position depending on the
verb's feature [tense], in Old Spanish, clitics enjoyed an actual syntactic freedom, as
shown in the contrast between the following pairs of sentences:
(7)
Asi commo lek el dezia. (Old Spanish)
as how to him-CL he said (*Modern Spanish)
As he said to him. «20) in Rivero 1986)
288
(8)
(9)
JON FRANCO
As! como eli lek dec!a. (Standard Modern Spanish)
, As he said to him.
Que ellos te
non digan en que
puede £lnar.
that they to you-CL not say in what can-3SG end
Let them not tell you how it could end.
(10) Que ellos no te digan en que puede terminar.
Let them not tell you how it could end.
10i al condei.
(11) Priso
Catch- PAST-3SG him-CL the count
He caught the count.
(12) Lo apres6 al conde.
He caught the count.
(Old Spanish)
(*Modern Spanish)
«13) in Rivero)
(Standard Mod. Sp)
(Old Spanish, (31) in Rivero)
(*Modern Spanish)
(Standard Modern Spanish)
,Rivero (1986) accounts for the phenomenon of the syntactic freedom of Old
Spanish clitics by showing that these clitics were syntactically equivalent to stressed
pronominals generated in A-positions. In other words, their behavior as clitics was
restricted to the phonological level in which clitics have to be bound to a stressed
'
vowel. 1
The obligatory adjacency to the verb of Modern Spanish clitics is even more
suspicious when examined cross-linguistically. Normally, the position of clitics is
syntactically constrained, but seldom lexically constrained. In Yagua, for instance,
Everett (1989) points out that object clitics can be affixed to any constituent as long
as the clitic is minimally c-commanding _its double. In Czech (Comrie 1989), the
object clitic has to be positioned after the first constituent of the clause, regardless of
its lexical nature.
Finally, another characteristic that clearly differentiates Modern Spanish from
Old Spanish is the obligatoriness versus the non-obligatoriness of clitic doubling
with overt object pronouns. The non-clitic version of sentences (13) and (14) below
actually occurred in Old Spanish «6) and (7) in Rivero), which somehow advocates
for a different status of object clitics in these two periods of the language:
o
(Standard Spanish)
(13) Dejad *(me) a mt.
Let me.
(14) Lo que a m! *(me) parece.
What seems to me.
(Standard Spanish)
0
In sum, the difference in behavior between Old Spanish and Modern Spanish
clitics with respect to clitic doubling with pronouns, the position of clitics, and
the types of constituents clitics cliticize to can be best accounted for by an analysis
that deems the diachronic development of clitics as a, transition from pronominal
(1) This makes elitic doubled constituents in Old Spanish non-arguments. Rivero (1986: 776) states in this
regard: "asp. doubling constructions have pronominal phrases (e.g. elitics) in subcategorized positions~ and
doubling phrases as base-generated adjuntcs of maximal projections".
CONDITIONS ON CLITIC DOUBLING: THE AGREEMENT HYPOTHESIS
289
elements to verbal agreement morphemes. The fact that Spanish clitics evolved from
Latin demonstrative pronouns might have preconized the commonplace grammatical
description of today's Spanish clitics as pronominal elements. In strict synchronic terms,
however, a pronominal analysis of Spanish clitics can be overridden by the descriptive superiority of the Agreement Hypothesis, as I will try to show in what follows.
2. Default object verb agreement in the paradigm
In answer to the first question raised in the introduction, I defend the idea that
there is a zero object agreement morpheme which is responsible for the lack of
phonologically overt agreement with certain object nominals. Justification for this
zero morpheme might be hard to draw at a first blush, though plausible if we take a
closer look at agreement relations crosslinguistically.
From a study of verbal agreement in 26 languages, Paus (1990: 29) concludes:
'The correlation of zero marking with third person categories, especially third ·person singular, is striking. It appears that in double agreement languages, there is a
universal tendency to treat third person as the unmarked person category. Third
person contrasts with first and second person which are almost always expressed by
means of an overt verbal affix.' In fact, in 20 out of the 26 languages in Paus'
sample, the third person was unmarked, suggesting a paradigmatic default zero
morpheme, as exemplified in (4) and (5) above.
Moreover, many languages make more finely grained distinctions, and make the
cut overt vs. non-overt marking between pronouns and noun phrases as in Welsh
(Doron 1988) or between animate noun phrases and inanimate ones as in Eshtehardi
object gender marking on the verb (Comrie 1989). Some lefsta2 dialects of Spanish, as
for instance Basque Spanish (Landa 1990), belong to this second group of languages.
Be that as it may, the division in the agreement patterns is often mapped onto an
animacy hierarchy whose highest positions are occupied by first and second person
pronouns, followed by third person pronouns then, by animate nouns, ... and lastly
by inanimate indefinite plural nouns (see Comrie 1989 and Silverstein 1976 for
further details on the hierarchy).
A· hypothesis that can be drawn from all these data is that the categories that
occupy a lower position in the animacy hierarchy may be defective categories with
respect to the presence of certain features, as in Cole (1991). According to Cole,
inanimate nouns in Korean and Chinese are defective since they lack the abstract
feature [antecedent a] which allows cat~gories to antecede reflexives. In parallel
fashion, one can assume that in Mansi (Paus 1990), for instance," and maybe in
Southern Cone Spanish as well, inanimate indefinite nominal objects are defective
categories since they take a zero marking on the verb. It would be plausible to
expect there to be crosslinguistic and crossdialectal variation as regards the specific
categories in the lower parts of the hierarchy that can be defective.
"Going back to the first question, it is legitimate to claim that certain varieties of
.(2) Leismo is the phenomenon that replaces accusative clitics by dative ones in contexts in which the so-cailed
etymological system requires accusative Case. There are several degress of leismo, the one described in Landa (1990)
affects only animate accusatives.
290
JONFRANCO
Spanish have a uniform paradigmatic object-verb agreement under the hypothesis
that a zero morpheme represents certain third person nominals on the verb agreement and the feature [3rd person] is precisely its only matrix value.
3. Object Agreement in Spanish.
In order to focus on the most challenging data, I am going to devote most of this
paper to the verb agreement of direct objects, since indirect objects, whether they
are nominal or pronominal, have never posed much problem as far as the objective
conjugation is concerned. First, clitic doubling with indirect objects is ,unrestricted
and, second, according, to several studies (cf. Silva-Corvahin 1981 and Bentivoglio
1978), the doubling, of indirect, objects by their clitics has become almost categorical.
We can see an example of this type of doubling in (15) below:
(15) Al
final ?(lei)
dije
la verdad a Pedroi.
At the end 3CL-OI told-1sg the truth to Pedro
At the end, I told Pedro the truth.
3.1. Clitic doubling, specificity and the Matching Principle
Silva-Corvahin (1984) and Sufier (1988), put up some good arguments and
empirical evidence in favor of the feature [+specific] as the main trigger for clitic
doubling in Southern Cone dialects. For instance, examples (16), (17) and (18) below
could be explained by saying that clitic doubling is possible when the doubled
element is [+specific]:
(Southern Cone Spanish)
(16)
LOi
conOCl al nuevo panaderoi.
CL-him met-I the new baker
I met the new baker.
(17)
Juan lai
saco
la
notai sin
Juan it-fern-CL got-3SG the
grade without
Juan got the grade without effort.
(18) *Juan lai
saco
una notai sin
Juan it-fern-Cl got-3SG a grade without
Juan got a grade without effort.
esfuerzo.
effort.
esfuerzo.
effort.
In particular, Sufier (1988) posits a Matching Principle which requires that
clitics - which are also treated as agreement, morphemes - should match the
features of the doubled NP. Moreover, according to her analysis, accusative clitics
are inherently marked [+specific] in the lexicon and subsequently the NP they
double must be also [ +specific] in order to set off the Matching Principle.
The first problem that this view entails is that one needs to demonstrate why
indirect object clitics are not inherently [+specific], since they may double [-specific]
IO-NPs. Otherwise, the argument turns out to be circular, that is, direct object
clitics are specific because they can only double [+specific] NPs and, only specific
NPs can occur in clitic-doubled DO constructions, because DO-CLs are [+specific].
291
CONDITIONS ON CLITIC DOUBLING: THE AGREEMENT HYPOTHESIS
Even though specificity seems to play a role in the clitic doubling phenomenon, I
believe it is the result of a side effect, that is, all elements that do not occupy a low
position in the Animacy Hierarchy are likely to be specific. Still, there is some
evidence that may shake the whole argument for the kinship between specificity and
clitic doubling. For instance:
(19) Juan 10i
invitaba
a unOi y
luego se
olvidaba.
Juan CL-him used to invite to one and then CL-3sg forgot
Juan used to invite people and then forget all about it.
(20)
En ese departamento, 10i
admiten
In that" department, CL-him admit,..3pl
In ~hat department, they admit anyone.
a cualquierai.
to anyone
(19) and (20) are cle~r examples of Do-clitic doubling with non-specific Nps. In
(19), a uno 'to one' can be analyzed as a generic pronoun (as suggested to me by
Zubizarreta, p.c.), and as in any object-verb agreement with a pronoun in Spanish, a
fully specified overt clitic is required. In example (20), however, a cualquiera 'to
anyone' is an animate noun and even though the nominal itself is not specified for
gender, the verbal agreement takes the masculine for this feature., "unless AGR
inherits the opposite value from a higher NP ala Borer (1989), as shown in (21):
(21) Si mi hermana fue admitida para enfermera,
if my sister
was accepted for nurse
iahf lai
admiten a cualquierai!
there CL-her accept-they to anyone
If my sister was accepted in the nursing program,
they accept anyone there!
The Matching Principle - although intuitively it might seem to be necessary at
some point - is, as stated in Sufler (1988), a rough way to account for agreement
facts. ,Agreement relations contain in most cases partial matchings of features as in
(21), and occasionally eve~ mismatches. Thus, a strong version of the Matching
Principle, that is, a one-to-one feature matching constraint between the clitic and its
double cannot be mantained if we are to account for the grammaticality of (20) and
(21). Furthermore, it is not clear at all what the formal status of the Matching
Principle is and how it could apply across the board.
Sufler (1988) also tackles the problem of the well-known contrast between DOclitic doubling and IO-clitic doubling with wh-elements in the light of specificity.
Consider the following asymmetry:
(22)
(23)
*I..A quieni
e1··';)
lOi
viste
who
3Acc-masc saw-you
Who did you see him/her?
iA quieni lei
hablaste ei por
who
3Dat talked-you by
Who did you talk to him/her on
telefono?
phone
the phone?
292
JONFRANCO
Along the lines of Sufier (1988), due to the fact that wh-words of the quien 'who'
and que 'whae type are [-specific] and since accusative clitics, such as 1o, are the only
ones inherently marked [+specific], there is in (22) a feature mismatching that
results in a violation of the Matching Principle. Nonetheless, the ungrammaticality
of sentences such as (22) and the like cannot be always understood as a specificity3
issue. For instance, Sufier gives the following example as ungrammatical:
(24)
*iA cuantas\cuales
pasajerasi
lasi
rescataron ei?
How many\which passengers-fern. CL-3pl-fem. rescued-3pl.
How many\which passengers did they rescue? «51b) in Sufier 1988)
Curiously, the a cutiles pasajeras 'which passengers' variant has a specific interpretation - though never admitted by Sufler (1988) - yet, the sentence is ruled out
by the grammar, contrary to our expectations.
3.2. The Agreement Hypothesis Analysis
Previous works that considered Spanish object clitics as agreement elements
generated in situ, such as Borer (1984) and Jaeggli (1982, 1986), treated clitics as
affixes attached to the verb. This approach, however, faces quite a problem, as noted
in Borer (1986), Borer & Grodzinsky (1986), and Saltarelli (1990). If the [V clitic +
verb] constituent is a lexical item, the Lexical Integrity Hypothesis of Chomsky
(1970) would impede this item to enter syntactic and morphosyntactic operations
such as agreement, giving way to undesired results.
In a development in the theory of Principles and Parameters, Pollock (1989) and
Chomsky (1989) decompose the IP node in several functional categories providing
us indirectly with a structural possibility to solve the problem mentioned above and
determine the canonical position of object clitics in Spanish. For this purpose, I will
adopt here a version of phrase markers similar to those of Chomsky's and Pollock's,
as illustrated in (25) below. In this way, assuming that the so-called object clitics are
verbal agreement morphemes in Spanish, I am going to pursue an analysis in which
object clitics are agreement heads (AGRo) that project agreement phrases (AGRP o).
Based on this analysis, I will attempt to derive the Spanish data discussed so far.
3.2.1. Preliminary considerations
Before we start deriving the relevant data, let us review the grammatical facts
involved. Verbal direct object agreement is obligatory with pronouns (in all dialects), truly optional with animate and definite inanimate nominals (in some dialects) and impossible with indefinite inanimate nouns and wh-words.
(3) Sufier's (1988) definition of specificity is quite vague: "Specificicy is used in this study co mean that
referent(s) of a [+spec] NP can be identified with a particular x in the linguistic context (ef Rivero 1977). This
definition of specificity as identifiability comes very close to some definitions on definiteness (e.g. Chafe 1976)",
(Sufier 1988: 397).
As regards sentence (24), it seems that the NP 'which passengers' can be identified with a particular subset of
people that belong to a specific set, i.e., the passengers involved in the wreck.
293
CONDITIONS ON CUTIC DOUBliNG: THE AGREEMENT HYPOTHESIS
(25)
AGRP s
SpnGR'
/~
AGR
TP
/~
Spec
T.
/~
T
AGRPo
/~
Spec
AGR'
/~
AGRo
VP,
/~
Spec
V'
/~
V
NPDO
NPIO
Agreement has several functions: it can have a referential function, an identification function, and a Case-assigning function. Normally one function does not exclude
the other, but on the other hand, not all agreement relations fulfill all these
functions. If I treat object-verb agreement morphemes as equivalent to subject-verb
agreement morphemes, the former should be able to assign some kind of objective
Case to the objects under the right configuration. However, if object-verb agreement can be absent sometimes we should infer that the verb has not lost its ability
to assign objective Case. These assumptions basically account for the optionality of
clitic doubling with certain types of nominals, already described. Thus, for some
Do-NPs the speaker can choose between the presence or absence of the verb-object
agreement morphology in the same way as slhe can choose between the passive and
active morphology.
3.2.2 Object Agreement and Accusative / Dative Case Assignment
Going back to the data, let us derive a sentence like (16) repeated below for
convenience as (26), which is an example of Do-clitic doubling or, in my approach,
verb-object agreement.
conoe! al nuevo panaderoi. (Southern Cone Spanish)
(26) (LOi)
CL-him met-I the new baker
I met the new baker.
The particle 10 'him' is base generated under AGRo and, as a morpheme, it
requires a host to be attached to. Therefore, Head-to-Head movement is triggered,
the verb moves to AGRo and adjuncts to it. At this point, in order for the DO al
nuevo panadero 'the new baker' to receive Case, it has to move to, Spec of VP and
subsequently to Spec of AGRPo since the verb has lost its power to assign Case. The
294
JONFRANCO
latter fact can be explained in two ways. A la Jaeggli (1982, 1986), the clitic or the
morpheme would absorb the verb's property to assign Case to its object. Alternatively, I will adopt a version of the Comp Indexing device, introduced in Aoun,
Hornstein and Sportische, (1981) to account for that-trace effects, and state that
when AGRo is visible at PF the features of the morpheme percolate up to the node
immediately dominating it and assign this node an index, as in (27). This index
thwarts the possibility for the verb to assign Case under government.
(27)
Once the direct object has landed on Spec of the AGRPo, Case assignment takes
place like in any other ordinary Spec-Head Agreement relation (cf. Sportische 1990).
Furthermore; it is at this stage of the agreement process that Sufier's Matching
Principle can be rescued in a weaker version that we will call the Feature Harmony
Principle, as stated in (28):
(28)
The Feature Harmony Principle. The feature matrixes of two elements
in an agreement r"elation - although they do not need to match
each pther exactly - have to be harmonious.
Somehow, the. Feature Harmony Principle is more promising than the Matching
Principle since the former is flexible enough to generate (26) as well as (20).
The spec of AGRo position is a very restricted one with respect to the types of
elements that it can host since it is an argument position selected by a head. Thus,
the head AGRo imposes selectional restrictions on its specifier in the same way that
verbs select their subjects. For the time being, let us say that an element with the
features [-definite, -animate] violates the selectional restrictions of AGRo in Southern Cone Spanish, hence it cannot appear as its specifier. The constraints that the
selectional restrictions of AGRo impose on the element in the Specifier position are
independent from those of the Feature Harmony Principle. That is to say, a sentence
can be well-formed according to the Feature Harmony Principle and still be ungrammatical if it does not meet the selectional restrictions of AGRo and viceversa.
Bearing this in mind and the fact that, featurewise, lower members of tpe
Animacy Hierarchy are defective, we can account for the following data in which
indirect object and direct object verb agreement are contrasted:
(29)
Lesi
di fuego
To them-CL
gave-I fire
I set some wood on fire.
a unas maderasi.
to some wood
VI
unas maderasi.
(30) *Lasi
them:"CL
saw-I some wood
I saw some wood.
The greater freedom of indirect object verbal agreement is derived straightforwardly from the way indirect objects receive Case in Spanish and Italian as well.
CONDITIONS ON CIITIC DOUBLING: THE AGREEMENT HYPOTHESIS
295
That is to say, since indirect objects obtain Case inherently as in Belletti and Rizzi
(1988) or lexically, they always occur with a preposition, they never have to move to
the specifier of any agreement head to receive structural Case. 4 Consequently, indirect objects do not need to fulfill the selectional restrictions of AGRo ; moreover,
they concord by virtue of a coindexing relation that abides by the Feature Harmony
Principle. On the other hand, (30) is ungrammatical because when the "defective"
DO unas maderas 'some wood' moves to Spec of AGR o to receive its Case, it becomes
subject to and a violator of the selectional restrictions of the verbal morpheme las
'them'. Likewise, one could account for the grammatical asymmetry between direct
object and indirect object wh-elements when entering verbal agreement relations, as
exemplified in (22) and (23) and repeated below as (31) and (32) respectively:
(31)
*~A quieni
10i
who
. 3Acc-masc
Who did you see?
(32)
viste
e 1··")
saw-you
i.A quieni lei
hablaste ei por telefono?
who
3Dat talked-you by phone
Who did you talk to on the phone?
However, to consider wh-words of the a quien 'to whom' type defective categories
for object agreement constitutes a bold stipulation which is difficult to motivate on
any ground. Alternatively, following our derivation of (26), let us assume that the
wh-element in (31) has moved to Spec of AGRo for Case reasons, and from there to
Spec of COMP leaving behind an empty category which is a variable. However, in
the flavor ofJaeggli (1986), this empty category is automatically identified as pro by
the features of the agreement morpheme or the clitic. This deprives the empty
category of its status as a variable. Thus, (31) can be ruled out because the empty
category left by the wh-element is a pro bound by an operator as in Jaeggli (1986), or
because the variable has lost its [-anaphoric] value. So far either solution seems to be
equally good. Parallel to (29), the grammaticality of (32) can be explained in terms
of the fact that the IO-wh-element has been extracted from within the VP and the
variable could never be identified by AGRo in that position.
Furthermore, the same results in grammaticality are reached with the wh-ele~ents in-situ in non-echo questions, a situation that seems to indicate that the
identification of empty categories by AGRo also holds at the LF component where
the wh-element moves to Spec of COMP to obtain scope. This is illustrated in (33)
and (34) below:
(33)
*iQuieni
IOk
vio
a quienk?
who
him-CL saw-3SG to whom
Who saw who?
(4) Actually, it would be more accurate to say (as pointed out to me by ]oseph Aoun) that the indirect object
cannot move to Spec of AGRo, that is, to a Case position, otherwise, it would violate Chomsky's (1986) Uniformity
Condition on inherent Case.
296
JON·FRANCO
(34)
iQuieni lek
dio
who
him-CL gave-3SG
Who gave a kiss to who?
un beso
a kiss
a quienk?
to whom
3.3. Zero object-verb agreement and obligatory overt object verb agreement·
Consider the following example:
(35)
Vimos
a Pedro.
saw-we
to Pedro
We saw Pedro.
Previously, J. have assumed that when the speaker selects the option of having
zero object agreement on the verb, as in (35) - although the object NP qualifies as
a non defective category - AGR:, contains only the feature third person since the
zero agreement option is not available with second and first persons. Then, since the
verb will eventually move up the tree to pick up its tense, Head-to-Head movement
applies and as a first step the verb adjoins to the object Agreement head. In the
instances represented by (35), AGRo is phonetically null and unable to assign an
index to the mother node. Hence, the Do-NP moves to spec of VP which is an
accessible position for the verb to assign it structural Case from the AGRo node.
The last set of data to be explained is that of object verb agreement with
pronouns as illustrated in (36):
(36)
No *(lai)'
VI
a ellai
por ningun sitio.
neg 3sg.fem saw to 3sg.fem anywhere '
I didn't see her anywhere. -
Due to the fact that object verb agreement is morphologically rich, one would
expect these morphemes to be able to license empty pro categories as the subject
verbal morphology does. In effect, this is the case, as shown in Jaeggli (1986) and
illustrated below:
(37)
Juan 10i
mato prOi.
Juan him-CL killed
Juan killed him.
(38)
* Juan mato proi.
Juan killed
Juan killed.
pro= him, Pedro,...
In (37), object agreement identifies the empty pronominal category, hence its
obligatoriness, abiding in this way by Rizzi's (1986) version of the Generalized
Empty Category Principle (39) and Jaeggli and Safir's (1989) Theory of Identification (40) which state that:
(39)
Let x be the licensing head of an occurrence of pro; then pro has the
grammatical specifications of the features on x coindexed with it.
(Rizzi 1986)
(40)
AGR can identify an empty category as thematic pro iff the category
containing AGR Case-governs the empty category.
(Jaeggli & Safir 1989)
CONDITIONS ON CLITIC DOUBLING: THE AGREEMENT HYPOTHESIS
297
However, from the point of view of an identification theory, the data in (36) are
puzzling since we have a fully specified overt pronoun that requires verb object
agreement obligatorily in order to be licensed. My stand at this point is that, as far as
verbal agreement is concerned, overt stressed pronouns in Spanish behave in the same
way as empty pros. Due to the pronominal nature of verbal agreement, both types of
elements are attracted to the specifier position of AGR where a healthy Agreement Case
Governed relationship is established (see Koopman 1990 in this regard).
Finally, it seems reasonable to ask why one cannot state that the AGRPo node is
not projected when the agreement morpheme is phonologically invisible. In my
analysis, AGRPo has to be projected every time there is a transitive verb. Additionally,
I also posited a third person zero object verbal agreement morpheme taking into
account that some clitic doubling dialects -of Spanish, e.g. Basque Spanish as in
Landa (1990), have the phenomenon of null objects in their syntax where neither the
object nor the clitic or agreement morpheme are present though always understood.
Therefore, if we want to maintain a Theory of Identification, it is necessary to have zero
morphemes in order to license these phonologicaly null arguments. Interestingly,
my analysis predicts that these null objects can only have as referents nominal
defective categories, mostly inanimate nominals, as it turns out to be the case in
Landa (1990).
4. Conclusion
This study argues not only for the feasibility of treating object clitics as object
agreement morphemes, but also for the superiority of such an analysis over the
pronominal-affix analysis. Potential descriptive problems in the literature of cliticdoubled constructions, such as asymmetries in wh-extractions, optional agreement and
ill-formed clitic-doubled object nominals are explained by the existence of an object
agreement phrase with clitics as heads, and the syntactic, semantic and subsequent
morphological restrictions that apply to the objects that move to the spec position of
this node. Furthermore, the availability of two accusative Case assigners, namely, the
verb and AGRo, allows us to have variation within the language.
References
Aoua, ]., N. Hornstein & D. Sportische, 1981, "Some aspects of Wide Scope
Quantification",]ournal ofLinguistic Research 1, 69-95.
Belletti, A. & L. Rizzi, 1988, "Psych Verbs and Theta Theory", NLLT 6,291-352.
Bentivoglio, P., 1978, "Formacian de cliticos: analisis sobre-el habla culta de Caracas", In H.
Lapez Morales (ed.), Corrientes actuales de la dialectologfa del Carihe Hispdnico, Editorial
Universitaria, Puerto Rico.
Borer, H., 1984, Parametric Syntax, Foris, Dordrecht.
- - - - , 1986, "Introduction", In Syntax and Semantics 19, Academic Press.
- - - , 1989, "Anaphoric AGR" , In O. ]aeggli & K.]. Safir (ed.), The Null Suhject
Parameter, Kluwer, Dordrecht, 69-110.
- - - - , & Y. Grodzinsky, 1986, "Syntactic Cliticization and Lexical Cliticization: The
Case of Hebrew Dative Clitics", In H. Borer (ed.) Syntax and Semantics 19,- Academic
Press, 175-217.
298
JONFRANCO
Chafe, W., 1976, uGivenness, Contrastiveness, Definiteness, Subjects, Topics and Point of
View", In C. Li (ed.), Subject and Topic, Academic Press, New York, 25-55.
Chomsky, N., 1970,' "Remarks on Nominalization", In R. Jacobs & P. Rosenbaum (ed.)
Readings in English Transformational Grammar, Mouton, The Hague, 184-221.
- - - - , 1986, Knowledge ofLanguage, Praeger, New York.
--'-----, 1989, "Some Notes on Economy of Derivation and Representation", In I. Laka &
A.K. Mahajan (ed.), MIT Working Papers in Linguistics 10, MIT Press, Boston, 43-74.
Cole, P., 1991, "Feature Percolation". Paper read at the Theoretical East..;Asian Linguistics
Workshop at VC Irvine,]an. 19.
Comrie, B., 1989, Language Universals and Linguistic Typology, The University of Chicago
Press', Chicago.
Craig, C., 1977, The Structure of]acaltec, University of Texas Press, Austin.
Doron, E., 1988, "On the Complementary of Subject and Subject-Verb Agreement", In M.
Barlow & Ch. A. Ferguson (ed.), Agreement in Natural Languages, CSLI, Stanford,
201-218.
Everett, D., 1989, "CUtic Doubling, Reflexives and Word Order Alternations in Yagua", Lg
65, 339-372.
Jaeggli, 0., 1982, Topics in Romance Syntax, Foris, Dordrecht.
.
, 1986, "Three Issues in the Theory of Clitics" In H. Borer (ed.), Syntax and
Semantics 19, Academic Press, New York, 15-42.
- - - - - & K. J. Safir, 1989, uThe Null Subject Parameter and Parametric Theory", In O.
Jaeggli & K. J. Safir (ed.), The Null Subject Parameter, Kluwer, Dordrecht, 1-44.
Koopman"H., 1990, ('Restrictions on Spec Position". UCLA ms.
Landa, A., 1990, ('Null Objects in Basque Spanish". USC ms.
Ortiz de Vrbina, J., 1989, Parameters in the Grammar ofBasque, Foris, Dordrecht.
Paus, C., 1990, "Subject and DO in Transitive Verb Agreement". USCms.
Pollock,J. Y., 1989, "Verb Movement, VG and the Structure ofIP", L1 20,365-424.
Rivas, A., 1977, A Theory ofClitics. Unpublished Ph.D. dissertation, MIT.
Rivero, M. L., 1986, uParameters in the Typology ofClitics in Romance and Old Spanish",
Lg 62,774-807.
Rizzi, L., 1986, "Null Objects in Italian and the Theory of pro",L1 17,507-557.
Saltarelli, M., 1990, ULa Gramatica de los Cl{ticos y sus Panimetros: Concordancia y Cadenas
Funcionales". Paper read at the Asociaci6n de Lingi.i{stica y Filolog{a de America Latina.
. IX Congreso Internacional, August 1990.
Silva-Corvahin, ·,C., 1981, I(The Diffusion of Object-Verb Agreement in Spanish", In H.
Contreras '&, J. Klausenburger(ed), Papers in Romance, Proceedings of the Tenth
.A:nniversary·Symposium on Romance Linguistics, Supplement 11,163-176.
- - - ' - , 1984, "Semantic and pragmatic factors in syntactic change", In). Fisiak (ed.),
Historical Syntq,x, ~outpn, Amsterdam, 555-573.
Silyerst~in, M., 1976, "Hierarchy of features and ergativity", In R. M. Dixon (ed.),
.Grammatical categories in Australian languages, Linguistic Series 22, Australian Institute
of Aboriginal Studies, Camberra, 112-171.
Sportiche, D., 1990, "Movement, Agreement and Case UCLA ms.
Strozer,)., 1976, Clitics in Spanish. Unpublished Ph.D. dissertation, UCLA.
Suiier, M., 1988, uThe role of AGR(eement) in Clitic-Doubled Constructions", NLLT 6,
391-434.
u
•